Gradient Dissent - A Machine Learning Podcast avatar

Dexa/Gradient Dissent - A Machine Learning Podcast

Learn more
Published May 20, 2021

Alyssa Simpson Rochwerger — Responsible ML in the Real World

Alyssa Simpson Rochwerger delves into the real-world applications of responsible machine learning, sharing her experiences in teamwork, the equitable COVID-19 vaccine rollout in California, and tackling healthcare data intricacies while emphasizing the ethical challenges and bias monitoring crucial for ethical AI development.
Episode Highlights
Gradient Dissent - A Machine Learning Podcast logo

Topics covered

Popular Clips

Episode Highlights

  • Bias Monitoring

    Alyssa Simpson Rochwerger emphasizes the importance of proactive bias monitoring in machine learning models. She advises setting up systems to capture real-time data and manually review it to identify biases, such as gender or racial bias, that could affect model performance 1. Alyssa shares a personal anecdote from her time at IBM, where a visual recognition system tagged an image of a person in a wheelchair as "loser," highlighting the need for thorough data review 2.

    It's appropriate and quite feasible to think critically around where you're getting your data and does it reflect the community or the people that you are going to be serving with the model?

    ---

    She stresses the importance of collecting diverse data to ensure models serve all intended populations effectively.

       

    Ethical Challenges

    Ethical and legal challenges often arise when deploying AI products, as Alyssa notes through examples like Amazon's biased hiring model. She explains that despite a model's technical success, it can be blocked by HR or legal teams if it doesn't align with broader organizational goals, such as diversity 3. Alyssa also discusses the need for a diverse team to navigate these challenges, emphasizing the importance of involving various departments early in the process 4.

    You need a whole team of people that is responsible for actually deploying something into a production environment.

    ---

    This approach helps in anticipating potential blockers and ensuring a smoother deployment.

       

    Real World AI

    Alyssa's book, "Real World AI," offers insights into the practical aspects of AI implementation, focusing on ethics and responsible AI. She shares her experience of writing the book, highlighting the collaborative effort involved and the challenges of gathering real-world stories about AI deployment 5. Despite the difficulties, Alyssa found the machine learning community eager to share their experiences, which enriched the book's content.

    People are really nice and they want to share their stories and they want to help others.

    ---

    Her work underscores the importance of transparency and collaboration in advancing responsible AI practices.

Related Episodes